microsoft / MaskFlownet

[CVPR 2020, Oral] MaskFlownet: Asymmetric Feature Matching with Learnable Occlusion Mask

Home Page:https://arxiv.org/abs/2003.10955

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

confusing about FlyingThings3d dataset preload

lelelexxx opened this issue · comments

Hi!
Thanks for sharing!
When I read the code, I noticed that, seems all the training data is preloaded into RAM? and
when training on FlyingThings3d dataset, you have separate it into several parts, and only load one part of it during training. I didn't find the code about reload the rest parts, Could you please point it out for me ? Thanks again!

Hi, thanks for your interest.

Yes, we load all data into RAM, and, for FlyingThings3D, you can load only a part by setting -s if your RAM is not enough (by default, the code loads the complete dataset). We do not implement reloading the rest parts.

Hi, thanks for your interest.

Yes, we load all data into RAM, and, for FlyingThings3D, you can load only a part by setting -s if your RAM is not enough (by default, the code loads the complete dataset). We do not implement reloading the rest parts.

I’m wonder how long it takes for training after preload data into RAM?

Hi, thanks for your interest.
Yes, we load all data into RAM, and, for FlyingThings3D, you can load only a part by setting -s if your RAM is not enough (by default, the code loads the complete dataset). We do not implement reloading the rest parts.

I’m wonder how long it takes for training after preload data into RAM?

It depends on your device. When using 4 TITAN XP, it takes about two or three days for training on FlyingChairs.